UNIVERSIDADE DE SÃO PAULO INSTITUTO DE FÍSICA DE SÃO CARLOS

Transcription

1 UNIVERSIDADE DE SÃO PAULO INSTITUTO DE FÍSICA DE SÃO CARLOS arxiv: v1 [hep-ph] 13 Aug 2014 Willian Matioli Serenone Heavy-Quarkonium Potential with Input from Lattice Gauge Theory São Carlos 2014

2

3 Willian Matioli Serenone Heavy-Quarkonium Potential with Input from Lattice Gauge Theory Dissertation presented to the Graduate Program in Physics at the Instituto de Física de São Carlos, Universidade de São Paulo to obtain the degree of Master of Science. Concentration area: Fundamental Physics Advisor: Prof. Dr. Tereza Cristina da Rocha Mendes Corrected Version (Original version avaiable at the Unity that lodges the Program) São Carlos 2014

4 AUTORIZO A REPRODUÇÃO E DIVULGAÇÃO TOTAL OU PARCIAL DESTE TRABALHO, POR QUALQUER MEIO CONVENCIONAL OU ELETRÔNICO PARA FINS DE ESTUDO E PESQUISA, DESDE QUE CITADA A FONTE. Ficha catalográfica elaborada pelo Serviço de Biblioteca e Informação do IFSC, com os dados fornecidos pelo(a) autor(a) Matioli Serenone, Willian Heavy-quarkonium potential with input from lattice gauge theory / Willian Matioli Serenone; orientadora Tereza Cristina da Rocha Mendes - versão corrigida -- São Carlos, p. Dissertação (Mestrado - Programa de Pós-Graduação em Física Básica) -- Instituto de Física de São Carlos, Universidade de São Paulo, Quantum chromodynamics (QCD). 2. Lattice formulation. 3. Potential models. 4. Numerical methods. I. Cristina da Rocha Mendes, Tereza, orient. II. Título.

5 THIS PAGE WAS INTENTIONALLY LEFT BLANK

6

7 To my parents, who taught me the value of knowledge

8

9 ACKNOWLEDGEMENTS I thank my parents, who always supported me with words of encouragement and valuable advice, not only during this project, but throughout my academic life. Thanks to them, several times I gave up the idea of giving up. I would like to thank especially my supervisor Tereza Mendes for the guidance and support offered, as well for her huge patience with the silly mistakes of a student who is starting his academic life. Undoubtedly, I could not have finished this project without her guidance, advice and corrections. I am grateful for her believing in my potential and presenting me to wonderful opportunities that I could never imagine I would have and that proved essential for my formation. One of these opportunities was the DESY Summer Student Proggramme of 2012, at DESY-Zeuthen in Germany. During those two months in the mid-2012 I learned key concepts for this dissertation and I am really grateful for the opportunity, as well as for the work and dedication of Dr. Karl Jansen who was my supervisor at the time. I thank the collaboration of Aleksandra Słapik during this time, who helped me at the developed project and became a friend. I would like to thank Cesar Uliana, Prof. Dr. Attilio Cucchieri and Dr. Benoît Blossier, for taking the time to read through this work, giving me useful suggestions to the text and collaborating to its precision. I also thank the staff of the the Instituto de Física de São Carlos (IFSC) for helping me countless times with the paperwork necessary to carry out my research. I am grateful to the professors of IFSC for increasing my knowledge in several aspects of physics. This research was funded by a FAPESP (São Paulo Research Foundation) fellowship, grant #2012/ Initial funding from CNPq is also acknowledged. I especially thank CRInt/IFSC for financial support, allowing my participation in the DESY Summer Student Programme mentioned above and in the Lattice 2013 Conference.

10

11 All things being consider d, it seems probable to me, that God in the Beginning form d Matter in solid, massy, hard, impenetrable, moveable Particles of such Sizes and Figures, and with such other Properties, and in such Proportion to Space, as most conduced to the End for which he form d them; and that these primitive Particles, being solids, are incomparably harder than any porous Bodies compounded of them; even so very hard as never to wear or break in pieces; no ordinary Power being able to divide what God Himself made one in the first Creation. Isaac Newton, Opticks Music happens to be a case of Artificial Quantization... As a result of the effective quantization, all significant descriptions of musical phenomena (... ) end up being expressed as dimensionless ratios between integers,..., henceforth known as pythagoreanisms. Pythagoras is known to have conjectured that it should be possible to express the whole of physical science as pythagoreanisms... Considering that in truth the world is quantized... we have to concede that Pythagoras guess was a real hit. Yuval Ne eman, Symmetry and "Magic" Numbers or From the Pythagoreans to Eugene Wigner Proceedings of the Wigner Centennial Conference The fact that, at least indirectly, one can actually see a single elementary particle in a cloud chamber, say, or a bubble chamber supports the view that the smallest units of matter are real physical objects, existing in the same sense that stones or flowers do. Werner Heisenberg, The Physicist s Conception of Nature

12

13 Abstract WILLIAN, M. S. Heavy-quarkonium potential with input from lattice gauge theory p. Dissertation (Master of Science) - Instituto de Física de São Carlos, Universidade de São Paulo, São Carlos, In this dissertation we study potential models incorporating a nonperturbative propagator obtained from lattice simulations of a pure gauge theory. Initially we review general aspects of gauge theories, the principles of the lattice formulation of quantum chromodynamics (QCD) and some properties of heavy quarkonia, i.e. bound states of a heavy quark and its antiquark. As an illustration of Monte Carlo simulations of lattice models, we present applications in the case of the harmonic oscillator and SU(2) gauge theory. We then study the effect of using a gluon propagator from lattice simulations of pure SU(2) theory as an input in a potential model for the description of quarkonium, in the case of bottomonium and charmonium. We use, in both cases, a numerical approach to evaluate masses of quarkonium states. The resulting spectra are compared to calculations using the Coulomb plus linear (or Cornell) potential. Keywords: Quantum chromodynamics (QCD). Lattice formulation. Numerical methods. Potential models.

14

15 Resumo WILLIAN, M. S. Potencial de quarks pesados com input de teorias de gauge na rede p. Dissertação (Mestrado em Ciências) - Instituto de Física de São Carlos, Universidade de São Paulo, São Carlos, Nesta dissertação estudamos modelos de potenciais com a incorporação de um propagador não-perturbativo obtido através de simulações de rede para uma teoria de gauge pura. Inicialmente fazemos uma revisão de aspectos gerais de teorias de gauge, dos príncipios da formulação de rede da cromodinâmica quântica (QCD) e de algumas propriedades de quarkônios pesados, i.e. estados ligados de um quark pesado e seu antiquark. Como um exemplo de simulações de Monte Carlo de modelos de rede, apresentamos aplicações nos casos do oscilador harmônico e teorias de gauge SU(2). Passamos então ao estudo do efeito de usar um propagador de glúon de simulações na rede como input em um modelo de potencial para a descrição do quarkônio, no caso do botômomio e do charmônio. Nós usamos, em ambos os casos, uma abordagem numérica para calcular as massas dos estados de quarkônio. Os espectros resultantes são comparados com cálculos usando o potencial de Coulomb mais linear (ou potencial Cornell). Palavras-chave: Cromodinâmica quântica (QCD). Formulação na rede. Modelos de potencial. Métodos numéricos.

16

17 List of Figures Figure 2.1 The process of discretizing the time. In our case x 0 x I, x N x F Figure 3.1 Visual representation of the association of a group element U µ (x) to a link of the lattice Figure 3.2 A graphical representation of a plaquette Figure 3.3 The process of integrating a paired link merge the two plaquettes forming a loop. In the left panel, two adjacent plaquettes with one paired link. In the right panel, plaquettes merged due to the integration Figure 3.4 Two loops with more than one paired link and the result of integrating over one of them (in red). Integrating over the other one then generates a trivial factor 1, as shown in Eq Figure 3.5 The process of tiling a 5 4 Wilson loop. In the left panel, all links are paired with plaquettes links. In the right panel, the loop is completely tiled 64 Figure 3.6 P as function of Monte Carlo time steps (t MC ). Notice that we can consider the system to be thermalized after 100 steps Figure 3.7 Results of P as a function of β in a simulation of a pure SU(2) gauge theory on a lattice of size Error bars are omitted due to their negligible size Figure 4.1 Feynman diagrams leading to the Coulomb potential in QED Figure 4.2 Poles of the lattice gluon propagator along the contours used for the evaluation of the integrals in Eq Figure 4.3 In red, the first term of the potential obtained from the lattice propagator (V LGP F 0 r). In blue, is the Coulomb-like potential (color factor included). 88 Figure 5.1 Comparison of wave functions for a Coulomb-like potential obtained through our computations with the analytic solutions

18 Figure 5.2 The mass spectrum of the bottomonium, along with the computed spectrum using the two different potentials and 8 states to fit the parameters (data from Table 5.4; see text for details) Figure 5.3 In the left panel, we compare the potential obtained using the lattice propagator (V LGP, in blue) with the Cornell potential (V Cornell, in red). The difference between these two potentials is shown in the right panel Figure 5.4 The computed spectrum using the two different potentials and 4 states to fit the parameters (data from Table 5.5; see text for details) Figure 5.5 Plot of wave functions for the bottomonium obtained through our computations, using the potential from the lattice propagator (in blue) and the Cornell potential (in red) Figure 5.6 The experimental mass spectrum of the charmonium, with the computed spectrum using the two different potentials (using data from Table 5.8). See explanation in the text Figure 5.7 Plot of charmonium wave functions obtained through our computations, using the potential from the lattice propagator (in blue) and the Cornell potential (in red). The left panel corresponds to S states and the right panel to P states. We use solid lines for states 1S and 1P and dashed lines for 2S and 2P Figure C.1 Continuum limit analysis for E Figure C.2 Continuum-limit analysis for E 1 using the two methods described in Section Figure C.3 x(0)x(τ) as function of τ for a =

19 List of Tables Table 5.1 Results for a Coulomb-like potential using two different step sizes to calculate the wave functions. The header of the first column contains the parameters used to generate the points to which we fit the results of our computation.. 95 Table 5.2 Naming scheme for bottomonium states Table 5.3 Experimental spectrum for the bottomonium and its preparation (see text) as input for our calculations Table 5.4 Comparison between the results obtained for the potential extracted using the lattice propagator and the usual Cornell potential, using 8 states to fit the parameters Table 5.5 Comparison between the results obtained for the potential extracted using the lattice propagator and the usual Cornell potential, using 4 states to fit the parameters Table 5.6 Naming scheme for charmonium states Table 5.7 Experimental spectrum for the charmonium and its its preparation as input for our calculations (see text) Table 5.8 Comparison between the results for the charmonium obtained for the potential extracted using the lattice propagator and the usual Cornell potential Table C.1 Results for E 0 using 10 5 Monte Carlo iterations, n = 10 and measurements made every five Monte Carlo iterations Table C.2 Results for E 0 using Monte Carlo steps, n = 10 and measurements made every 25 Monte Carlo iterations Table C.3 Fitting parameters for functions in Fig. C Table C.4 Results for E 1 using Monte Carlo iterations, n = 10 and measurements made every each 25 Monte Carlo iterations Table C.5 Results for E 1 using Monte Carlo iterations, n = 5 and measurements made only every 50 Monte Carlo iterations Table C.6 Parameters used for the fit of the graph in Fig. C

20

21 Contents 1 Introduction 21 2 Overview of Gauge Field Theories Feynman s Path Integral Connection with Statistical Mechanics Abelian Gauge Fields Non-Abelian Theory Fermions Euclidean Action for Gauge Theories Quantum Gauge Field Theory Quantum Chromodynamics (QCD) Lattice QCD and Monte Carlo Simulations Lattice Gauge Theory Strong-Coupling Expansion Linearly Rising Potential Monte Carlo Methods Simulating a Gauge Theory with the Metropolis Algorithm Some Results Potential Models for Heavy Quarkonia Brief Review of Potential Models Potential from Lattice Propagator Method for Obtaining Quarkonium Masses Results Code Testing Bottomonium Charmonium

22 5.4 Summary Conclusion 113 References 117 A Used Notation 123 B Group Theory and Group Integration 127 B.1 Definition of a Group and Useful Concepts B.2 Lie Groups and Lie Algebras B.3 Example of Lie Group: The SO(3) Group B.4 Group Integration B.5 Example: The SU(2) Group C Monte Carlo Simulation of the Harmonic Oscillator 143 C.1 Analytic Solution with Path Integrals C.2 Numerical Results

23 21 Chapter1 Introduction The study of bound states has been important throughout the history of particle physics. In the beginning of the 20th century, the need to understand the atom structure was one of the challenges that led to quantum mechanics. When physicists started to probe the structure of matter at even smaller scales, they found out that the atomic nucleus was a bound state as well. Then, the need to understand how protons and neutrons bind together to form the nucleus led to the proposal of the strong force, later to be described by quantum chromodynamics (QCD). According to QCD, a quantum field theory with local gauge symmetry, the proton itself is a bound state of quarks. The study of bound states due to the strong force can be implemented directly from QCD, in a first principles approach, using the lattice formulation of gauge theory. It is interesting to note that the quark model, proposed in 1964, became widely accepted only in 1974, after the discovery of the J/ψ particle, which is a bound state of charm and anticharm quarks (1). In this dissertation, we consider bound states of a heavy quark and its antiquark in the nonrelativistic approximation, using input from lattice-gauge-theory calculations. Gauge theories are the foundation of the Standard Model of particle physics. Today the model includes six flavors of quarks, carrying electric charge +2/3 (quarks u, c, t) or 1/3 (quarks d, s, b). There are six leptons as well: e, µ, τ and their corresponding neutrinos ν e, ν µ, ν τ. These are the fermions of the standard model. The electromagnetic force is described in a unified way with the weak force, which is responsible for the decay of quarks and leptons. The bosons associated with the electroweak force are the

24 22 Introduction photon (carrier of the electromagnetic force) and the W ±, Z 0 particles, associated with the weak interaction. Since in the unified theory the W ± and Z 0 bosons are massless, conflicting with the experimental evidence, a new boson is introduced: the Higgs boson. The interaction of the particles with the Higgs field explains the origin of their masses (although no prediction is made for the mass values, except for the mediating bosons W ±, Z). The existence of the Higgs particle was confirmed in The bosons associated with QCD comprise eight color-charged gluons, the force carriers of the strong interaction, responsible for gluing the quarks. Gravity is not a part of the standard model, although the existence of the graviton, which would be the carrier boson of the gravitational force, is speculated. All the three forces of the standard model are described by gauge theories. Nevertheless, for a long time, the importance of gauge theories was neglected. During the first half of the 20th century, physicists regarded the gauge symmetry of electromagnetism as an accident. This accident was the only thing keeping gauge invariance alive (2). The importance of gauge symmetries was only recognized later, in the 1950 s, when quantum field theory was being developed. Even then, the description of the strong force by a gauge theory remained a challenge for about two decades. In the early 1930 s, quantum mechanics was well developed. The first steps toward a quantum field theory were being made, following the theoretical prediction of antiparticles by Dirac. This prediction was confirmed in 1932, with the discovery of the positron in cosmic rays. It was understood in that period that the electromagnetic force was mediated by the exchange of photons. Physicists of the time knew as well that an atom was made up of electrons, protons and neutrons, with the neutron also being discovered in However, this picture had unsolved issues. One of these issues was the beta decay, which was unexplained and required the conversion of a neutron into a proton. The process seemed to violate conservation of energy. This led to the proposal by Pauli in 1931 of another, much lighter, neutral particle. The idea was used by Fermi in 1934 to propose a new interaction, later called the weak force, which was responsible for the creation of the missing particle. This particle was named neutrino some years later (3). Another issue was: what could hold dozens of positively charged particles bound in a

25 23 small region of space such as the atomic nucleus? And why was the neutron there and did not wander outside the nucleus? The solution was to postulate the existence of an unknown strong force binding the nucleus together. Also, this force should be of short range, confined to the atomic nucleus. Under the picture of forces being mediated by the exchange of particles, Yukawa proposed in 1935 the existence of a massive particle that would be the strong-force equivalent of the photon in electromagnetism. Experimental confirmation came only in 1947 when Lattes, Muirhead, Occhialini and Powell discovered the π-meson (4). In the following years, several other strongly interacting particles, or hadrons, were discovered. The abundance of elementary particles made physicists wonder if all of them were truly elementary, or if they were different bound states of a handful of particles. Indeed, in 1964 Gell-Mann and Zweig independently proposed the existence of particles that were more elementary and composed the proton, neutron, pion and the other recently discovered hadrons. Gell-Mann named them quarks. In the late 1960s, the observation of high-energy electron scattering in a proton showed that the proton behavior was indeed that of a bound state of three point-like particles (3). Despite this, the quark model was not well accepted until the so-called November revolution in 1974, with the discovery of the J/ψ meson. A compelling reason not to believe in quarks was that they were never observed in isolation. Also, the model predicted that quarks were 1/2-spin particles. But, for some hadrons such as the ++ particle, three quarks would have to be in the same quantum state, thus conflicting with Pauli s exclusion principle. The solution came from the hypothesis that these particles had a new quantum number a new charge, the color charge which could be of three different types: red, green or blue. The fact that they had different quantum numbers allowed them to occupy the same state without violating Pauli s principle. These new charges would be the source of the strong interaction. The force binds the quarks inside hadrons in such a way that the bound states are all colorless (neutral, or white). More precisely, the three quarks inside a baryon are of different colors, resulting in a colorless state, or the quark and antiquark in a meson carry a color and its respective anticolor. As a consequence, the color charge cannot be observed. It is confined inside the hadrons.

26 24 Introduction The theory of the strong force is called quantum chromodynamics, due to the analogy with color combinations. Since quantum electrodynamics (QED) had yielded excellent results and was unified with the weak force by Weinberg and Salam in , it was natural to suppose that QCD would be a gauge theory as well. However, the conventional treatment presented problems in the QCD case, since the theory was predicted to have a diverging coupling constant at physically relevant short distances. The issue was solved in 1973 by Gross, Politzer and Wilczek (5, 6), who showed that the theory s coupling tends to zero in the limit of high energy (short distances), in accord with the deep-inelastic-scattering experiments mentioned above, which showed that the quarks behave as free particles at high energies. This phenomenon is called asymptotic freedom. Conversely, this behavior suggested that the coupling constant could be strong at low energies (long distances), which would account for the confinement of quarks. It is important to note that the methods used to study QED cannot be applied in the low-energy regime of QCD, since they rely mainly on the fact that the QED coupling parameter, the fine-structure constant, is much smaller than one, making the theory suitable for perturbation theory. In fact, asymptotic freedom tells us that perturbation theory is only applicable to QCD at high energies. In other words, the nature of the QCD coupling, which is small at high energies and large in the low-energy regime, determines that QCD bound states, such as the proton, cannot be described perturbatively. A method to perform non-perturbative computations in QCD came in 1974 when Wilson showed how to discretize the theory on a lattice, retaining its gauge symmetry exactly (7). This enabled him to perform a strong-coupling expansion to calculate pathintegral averages, using a similar method to the high-temperature expansion of classical statistical mechanics. In particular, Wilson could prove confinement of static quarks by a linear potential in the strong-coupling limit, i.e. when the lattice parameter β is small. In order to extend this result to the physical limit, however, it is necessary to consider large values of the lattice parameter, which was not possible using the strong-coupling expansion. The lattice formulation was later used in computational simulations to obtain nonperturbative expectation values of QCD observables at physical values of the coupling.

27 25 It was then possible to show that, in the limiting case of infinitely massive quarks, the interaction potential between a quark and its respective antiquark rises linearly (8). Bound states composed of a quark-antiquark pair are called quarkonia. The consideration of the limit of heavy quarks (charm, bottom) in QCD offers the opportunity of a more direct approach to several fundamental aspects of the theory, such as probing perturbative effects on the gluon fields (9). Also, the study of heavy quarkonia may be done in a quite precise way by nonrelativistic potential models (10). This is possible because in this case the binding energy becomes much smaller than the quark mass, justifying the nonrelativistic approximation. In this sense, these states may be viewed as the positronium of QCD, since we can try to obtain the system s several observed states using Schrödinger s equation. The method allows the study of all the energy spectra, while enabling an explicit physical interpretation of the interactions between the particles in the bound state. The potential should be modeled with respect to the physical characteristics of the QCD running coupling, described above. Generally, the potential is a sum of two terms: The first one, obtained perturbatively, comes from the quark-antiquark interaction in the approximation of one-gluon exchange (OGE). It can be related to elastic scattering inside the meson [see (11, Chap. 6)]. The second one is a linearly-rising confining potential, inspired by lattice calculations. The resulting potential is called Coulomb-plus-linear potential or Cornell potential V (r) = 4 α s 3 r + F 0 r, (1.1) where α s and F 0 are suitable constants. A list of common potentials is given in Ref. (10). The binding energy of the quark-antiquark pair may be obtained by a numerical integration of the Schrödinger equation, yielding the mass spectrum of the system. Alternatively, the contribution to the potential coming from the OGE term cited above may be obtained directly from the theory s gluon propagator. This is what we propose in our study, using data for this propagator obtained from lattice simulations of the theory [see e.g. (12)]. In Chapter 2 of this dissertation we review the path-integral formalism in gauge theories, both in the Abelian and the non-abelian case. Also, we discuss the quantization

28 26 Introduction of gauge theories using path integrals and comment on general features of QCD. We proceed in Chapter 3 to discuss the discretization of QCD on a lattice, which enabled the strong-coupling expansion mentioned above and justifies the linearly rising potential that we will be using later. The lattice discretization is well suited to computer simulations of the theory, using Monte Carlo methods. As an application, we developed a program to simulate the case of pure SU(2) gauge theory (i.e. without dynamical fermions) and present some of our results. Chapter 4 details the calculation of the OGE contribution to the potential as mentioned above and describes the numerical method we use to find the eigenenergies in the Schrödinger equation. Our results, incorporating the nonperturbative gluon propagator into the OGE term of the potential, are presented and discussed in Chapter 5. Our conclusions are drawn in Chapter 6. Appendix A summarizes the used notation, while Appendix B reviews some aspects of group theory.

29 27 Chapter2 Overview of Gauge Field Theories Don t turn your back. Don t look away. And don t blink. Good Luck. The Tenth Doctor Doctor Who: Blink In this chapter we review some basic aspects of gauge theory. Firstly, we introduce the Feynman path-integral formalism, showing that it is equivalent to the operator formalism of nonrelativistic quantum mechanics (13). It is possible to show that the path-integral formalism leads to Hamilton s principle in the classical limit. The same procedure applies to quantum field theories (14). We start by briefly reviewing the case of electromagnetism, emphasizing its Abelian symmetry. Then we build on top of it a gauge theory with non- Abelian symmetry (2, 15). We also discuss the inclusion of fermions. Finally, we perform the quantization of the classical gauge theory through the path-integral formalism. This provided a framework to study quantum field theory by perturbation theory, based on an expansion in Feynman diagrams, which was remarkably successful for QED. This treatment is applicable to QCD only in the high-energy regime, where the property of asymptotic freedom holds. We note that, due to the way in which it is defined, the path integral also provides a natural connection of quantum field theories with statistical mechanics (16). In fact, it is one of the main ingredients of the lattice formulation of gauge theories (15), which allows the nonperturbative investigation of QCD using statistical mechanical methods, such as

30 28 Overview of Gauge Field Theories the strong-coupling expansion and Monte Carlo simulations. This will be instrumental in Chapter Feynman s Path Integral Our aim here is to derive the Feynman path integral from the usual operator formulation used in nonrelativistic quantum mechanics (14, 13). For simplicity, let us consider a one-dimensional system. We start by calculating the probability amplitude of a particle leaving a position x I to arrive at a position x F under the action of a potential V (x) during a time T. The probability is given a by x F e iht x I. We divide the time T in N steps of size a, so that Na = T (see Fig. 2.1). This will result in x F e iht x I = x F e } iha e iha {{... e iha } x I. (2.1) N factors Figure 2.1 The process of discretizing the time. In our case x 0 x I, x N x F. x 0 x 1... a x N 1 x N Source: By the author. We then insert the completeness relation dx x x = 1 between each two consecutive exponentials, to get x F e iht x I = ( N 1 i=1 dx i ) x F e iha x N 1 x N 1 e iha x N 2... x 2 e iha x 1 x 1 e iha x I. (2.2) The interpretation of the above expression is that we compute the probability amplitude for the particle leaving a position x i at time t i to reach the position x i+1 at time t i+1, for all the N time slices. The next step is to multiply these amplitudes to obtain the probability amplitude for the particle leaving x I to arrive at x F in a time T along a a We will adopt natural units for all our calculations, see Appendix A.

31 2.1 Feynman s Path Integral 29 specific path. We repeat this for all possible paths and sum the corresponding amplitudes to obtain the total probability amplitude x F e iht x I. Let us proceed then to compute x i+1 e iha x i. The Hamiltonian for a particle of mass m is written as H = K(p) + V (x), corresponding to kinetic and potential terms respectively. Since we plan to take the limit a 0 later, we may approximate e iak(p) iav (x) = e iak(p) e iav (x) e a2 2 [K,V ]... e iak(p) e iav (x), (2.3) where we retained only the leading contribution in the Zassenhaus formula [see e.g. Eq. 1.2 of Ref. (17)]. Next, we use Eq. 2.3 and the completeness relation in the momentum basis dp p p = 2π to write x i+1 e iha x i as x i+1 e iha x i x i+1 e iak(p) e iav (x) x i = e iav (x i) dp 2π x i+1 e ia p 2 2m p p xi. (2.4) We then use the relation x p = e ipx to obtain a Gaussian integral. We get x i+1 e iha x i = e iav (x ] i) dp exp [ ia p2 2π 2m + ip(x i+1 x i ), (2.5) which can be easily evaluated by completing squares ( ) { [ 1/2 ( ) 2 im x i+1 e iha m xi+1 x i x i = exp ia V (x i )]}. (2.6) 2πa 2 a Now we have the amplitude for a particle leaving position x i to arrive at position x i+1 in a time a. We can insert Eq. 2.6 into Eq. 2.4 to obtain the total probability amplitude x F e iht x I = ( ) N 1 im 2 2πa ( N 1 i=1 dx i ) exp { ia N 1 j=0 [ m 2 ( ) 2 xj+1 x j V (x j )]}. (2.7) a But time is not a discrete quantity. This means that we must consider now the continuum limit, making a progressively smaller (and, conversely, N progressively larger). In this limit we define ( ) N 1 im 2 2πa ( N 1 i=1 dx i ) D[x(t)],

32 30 Overview of Gauge Field Theories x j+1 x j a N 1 a j=0 ẋ, dt, (2.8) to obtain the final expression for the probability amplitude x F e iht x I = { D[x(t)] exp i [ mẋ 2 dt 2 ]} V (x). (2.9) We can see that the integral in the exponential is just the definition of the classical action. Finally, we rewrite the probability amplitude as x F e iht x I D[x(t)] e is[x(t)]. (2.10) The inverse process, where we start from the path integral formulation and show its equivalence with the Schrödinger equation, is shown e.g. in Ref. (18). Let us notice that this formulation keeps a strong connection with classical physics. In fact, we are using the classical action to perform the calculations. It is also possible to show, as a limiting case, that one recovers Hamilton s principle (i.e. δs = 0) in the classical regime (13). To see this, we need to reintroduce Planck s constant into Eq x F e iht/ x I = D[x(t)] e is[x(t)]/. (2.11) Now we treat as a parameter, for which the limit 0 can be taken. Considering neighboring paths, the action S[x(t)] varies slightly. However, in units of this variation is enormous, so that the phase e is[x(t)]/ will be a rapidly varying (periodic) function. As a consequence, the contributions of these paths will cancel each other in the sum and the corresponding probability amplitude will be zero. The only nonzero contribution comes from a path corresponding to an extremum of the phase, i.e. the one for which the phase is stationary. This is the same as requesting that the action be extremized and therefore we recover Hamilton s principle. An alternative to this qualitative argument is described in Ref. (14).

33 2.2 Connection with Statistical Mechanics Connection with Statistical Mechanics The path integral has an analytic solution only in some simple cases. Also, in the form in Eq. 2.10, it is not efficient to evaluate it numerically. Our aim here will be to relate the path integral to the partition function in statistical mechanics, which will enable us to use statistical mechanical techniques to address the problem. In particular, as will be discussed in Chap. 3, we may use Monte Carlo simulations to evaluate the discretized version of the path integral. We transform from the Minkowski time t to the Euclidean time τ = it. The names come from the fact that, when we do this transformation, the Minkowski metric is replaced by the Euclidean metric (see Appendix A). This procedure is known as Wick rotation (19). We get Z F I x F e iht x I { = D[x(τ)] exp dτ [ m 2 ( ) 2 dx V (x)]} dτ D[x(τ)] e S[x(τ)], (2.12) where S[x(τ)] is the Euclidean action. Notice that, in Euclidean time, the Lagrangian assumes the expression of the Hamiltonian in Minkowski time. Also, the path integral is now extremely similar to the partition function in statistical mechanics. This means that we can use statistical mechanical methods to evaluate it and to obtain other observable quantities such as the ground-state energy of the system. Before proceeding, we define the average of an operator  as ( ) Tr e HT  x e HT   x dx =, (2.13) Z Z where Z is defined by [see Ref. (16)] Z Tr Z F I = dx x e HT x = with Z F I given in Eq dx I dx F Z F I δ(x F x I ), (2.14) We remark that  stands for a statistical average, while x  x is the quantum mechanical expectation value. Note also that Z is just Z F I with x F = x I and integrated

34 32 Overview of Gauge Field Theories over all possible values for initial (or final) points. Thus it is a path integral. By the same reasoning, the average  in Eq is a path integral as well. We can write it as  = D[x(τ)] A[x(τ)] e S[x(τ)] D[x(τ)] e S[x(τ)], (2.15) where the Euclidean action S[x(τ)] is defined in Eq Note that we slightly changed our definition of path integral here, to include the end points of the paths. This means that the product defined for discrete time steps in Eq. 2.8 now runs from i = 0 until N, where x 0 = x I, x N = x F and x I = x F. We will use this notation from now on. If our system has M (M may be infinity) discrete energy levels, we can insert the completeness relation M 1 n=0 n n = 1 into Eq. 2.13, obtaining  = 1 Z M 1 HT x e n n  x dx = 1 n=0 Z M 1 n=0 EnT e n  n, (2.16) where n = 0 is the ground state. Also, we interchanged the sum and the integral, and used the completeness relation dx x x = 1. Taking the limit T, we find  0  0. (2.17) Note that the absence of the term e E0T comes from the fact that Z e E0T in this limit, leading to a cancellation. This can be easily verified by setting  = 1 in Eq In this way, we can isolate the ground energy level by studying the behavior of the Hamiltonian at large T, which means evaluating the path integral in Eq in this limit. For our calculations we make use of the virial theorem(18) mẋ 2 = xv (x), (2.18) which allows us to write Eq as E 0 = 0 H 0 = lim T Ĥ = D[x(t)] [xv (x)/2 + V (x)] e S[x(t)]. (2.19) D[x(t)] e S[x(t)] For the calculation of the first excited state, we introduce the time-ordered (in Eu-

35 2.2 Connection with Statistical Mechanics 33 clidean time) connected N-point function b Γ (N) c [( n i=1 ) δ ln Z(J)], (2.20) δj(τ i ) J=0 where J J(τ) is a function of the Euclidean time and represents an external source added to the partition function [see e.g. Ref. (20)], i.e. Z(J) generalizes Z in Eq to { [ ]} Z(J) = Tr exp HT + x(τ )J(τ )dτ. (2.21) Notice that in Eq we mean a functional derivative. In particular, δj(τ)/δj(τ i ) = δ(τ τ i ). Also, by J = 0 we mean to set J(τ) to zero at all times. If we choose N = 2 and τ 1 = τ, τ 2 = 0 to evaluate Eq. 2.20, we obtain Γ (2) c (τ) = Tr [ e HT x(τ)x(0) ] Tr [ e HT x(τ) ] Tr [ e HT x(0) ], (2.22) Z Z Z which gives, using Eq. 2.13, Γ (2) c (τ) = x(τ) x(0) x(τ) x(0). (2.23) We then consider the limit T, which allows us to use Eq. 2.17, and rewrite the expression above as Γ (2) c (τ) = 0 x(τ) x(0) 0 0 x(τ) 0 0 x(0) 0. (2.24) By inserting the completeness relation n=0 n n = 1 between x(τ) and x(0), we find Γ (2) c (τ) = n 0 0 x(τ) n n x(0) 0. (2.25) We must highlight that the operators x(τ) and x(0) are in the Heisenberg representation. This means that there is a time dependence in the above equation. We can extract it by transforming to the Schrödinger representation through the relation x(τ) H = e Hτ x S e Hτ. We arrive then at Γ (2) c (τ) = 0 x n 2 e (En E0)τ. (2.26) n 0 b By time-ordered we mean that we will not be able to commute the time-dependent operators evaluated at different times. The times are ordered from higher to lower values of τ, respectively from left to right, i.e. τ 1 > τ 2 > > τ n.

36 34 Overview of Gauge Field Theories If we consider the limit of large τ, we actually manage to isolate the difference between the ground state and the first excited state Γ (2) c (τ) 0 x 1 2 e (E 1 E 0 )τ. (2.27) This approach is of central importance in numerical simulations, which are done based on the discretized version of Eq. 2.12, namely ( ) N 1 ( im 2 N 1 ) { Z F I = dx i exp a 2πa i=1 N 1 j=0 [ m 2 ( ) 2 xj+1 x j + V (x j )]}. (2.28) a In fact, all the above expressions for Γ (2) c (τ) carry over to discrete times, i.e. by taking x(τ) x i and J(τ) J i. The two-point function is obtained directly from the thermalized x i configurations generated in a Monte Carlo simulation. There are then two possibilities for determining E 1 from the computation of Γ (2) c (τ). The first one consists in plotting the results for Γ (2) c as a function of τ and then fitting to these points the function Eq This will work better if we consider the larger [ ] values of τ. The other possibility is to plot the quantity ln Γ (2) c (τ + τ)/γ (2) c (τ). This will remove the constant in front of the exponential in Eq and then we consider the expression [ E 1 E 0 = 1 τ ln Γ (2) c ] (τ + τ). (2.29) Γ (2) c (τ) This second method is more reliable since we look for a plateau for the quantity on the RHS of the above equation at large values of τ and read off E 1 E 0 directly. 2.3 Abelian Gauge Fields In this section we will briefly review the case of electromagnetism, emphasizing its Abelian symmetry (2). We postulate the existence of the vector potential A µ and then define the field-strength tensor as F µν = µ A ν ν A µ. (2.30)

37 2.3 Abelian Gauge Fields 35 We adopt the usual covariant notation, with µ = 0, 1, 2, 3 and Einstein sum convention for repeated indices. Details of the used notation can be found in Appendix A. The theory s Lagrangian density is given by L = 1 4 F µνf µν j µ A µ. (2.31) It is tempting to identify j µ with the electromagnetic-charge current density. However, we will refrain from doing this for a while. For now, it will be treated as an external current, as was done in Eq. 2.21, which may be useful in computing correlation functions of A µ (x). Later, in Section 2.5, we will be able to give physical meaning to it. We want to substitute Eq into the Euler-Lagrange equations [ ] δl δl µ = 0. (2.32) δa ν δ( µ A ν ) Noting that we obtain δ(j µ A µ ) δa ν = j ν, (2.33) δf λσ δ( µ A ν ) = δµ λ δν σ δ µ σδ ν λ, (2.34) j ν 1 4 [ ( µ (δ µ λ δν σ δ σδ µ λ) ν F λσ + F λσ δ λµ δ σν δ σµ δ λν)] = j ν 1 4 µ [(F µν F νµ ) + (F µν F νµ )] = j ν 1 2 µ [(F µν + F µν )] = 0, (2.35) where, in the last step, we used the fact that the field-strength tensor is antisymmetric with respect to its indices, which can be easily seen from its definition in Eq We thus have the equations of motion µ F µν = j ν, (2.36) which is the covariant form of the two Maxwell equations with sources, i.e. the Gauss law and the Maxwell-Ampère law. It remains to prove that F µν will obey the Gauss law for the magnetic field and Faraday s law. Let us notice that ρ F µν + ν F ρµ = ρ µ A ν ρ ν A µ + ν ρ A µ ν µ A ρ

38 36 Overview of Gauge Field Theories = µ ρ A ν µ ν A ρ = µ F νρ. (2.37) This leads us to the identity [see e.g. Ref. (2)] ρ F µν + ν F ρµ + µ F νρ = 0, (2.38) which accounts for the two missing laws. Note that, if we differentiate Eq with respect to ν, we obtain an antisymmetric expression on the LHS. The RHS, however, is not altered by the exchange of indices, which leads us to the continuity equation µ j µ = 0. We are interested in understanding the symmetry of the theory. By this we mean a transformation that will keep the equations of motion in Eq invariant. This transformation is called a gauge transformation and is written as A µ A µ = A µ + µ Λ(x), (2.39) where Λ(x) is an arbitrary function of the space coordinates. We say that the symmetry group of electromagnetism is the U(1) group. We can explicitly see this by representing a given group element U of U(1) by U = e ig 0Λ(x), (2.40) where g 0 is a coupling constant, which in the present case can be identified with the elementary electric charge. Writing U in this way effectively associates a group element to each point in space-time. Using this representation, we may write the gauge transformation as A µ A µ = UA µ U 1 i g 0 ( µ U)U 1. (2.41) Eq seems to be an unnecessary complication if compared to Eq However, we will see that this formulation will allow us to easily use electromagnetism as a basis for a theory with SU(N) gauge symmetry. Let us use Eq to see how the field-strength tensor F µν transforms under a gauge transformation. We get F µν F µν = µ A ν ν A µ = µ (A ν + ν Λ) ν (A µ + µ Λ) = F µν. (2.42)

39 2.4 Non-Abelian Theory 37 In this way we determine that F µν is invariant under a gauge transformation. Since j µ is an external current, we have the freedom to define its behavior under a gauge transformation. In particular, we may assume that j µ will transform in such a way that the term involving it will be kept invariant. When interpreting j µ as physical current in Section 2.5, we will be able to see that the term j µ A µ is not invariant c. Indeed, the extra term that appears is needed to keep the full Lagrangian (involving the gauge fields and fermions) invariant. Since F µν is invariant under gauge transformation, we get that the Lagrangian will be invariant as well L L = F µν F µν j µ A µ. (2.43) Also, notice that, since the gauge transformation does not affect F µν, Eq will be invariant as well. 2.4 Non-Abelian Theory We proceed to build a gauge theory with non-abelian symmetry (15). We perform the same calculations as above but considering a non-abelian gauge theory with SU(N) symmetry. As above, we will write the group elements as U = e ig 0Λ a (x)λ a, (2.44) where λ a are the group generators (see Appendix B) and g 0 is the bare coupling constant of the theory. The index a is called a color index and runs from 1 to N 2 1 (see Appendix A). The generators obey the relation [λ a, λ b ] = if abc λ c, (2.45) where the braces stands for the commutator. We adopt the normalization Tr(λ a λ b ) = 1 2 δab. (2.46) c Note that if we choose j µ = j µ, the term involving j µ will transform as j µ A µ j µ A µ + j µ µ Λ. Nevertheless, this additional term does not affect the equations of motion (since the additional term does not depend on A µ ) and, for now, no harm is done by imposing j µ A µ to be gauge-invariant.

40 38 Overview of Gauge Field Theories One important point to notice is that when we allow U to be a group element of SU(N) the vector potential becomes a non-commuting object as well, which can be represented in matrix form. We can decompose it as a linear combination of the SU(N) generators λ a A µ = A a µ λ a. (2.47) We can invert this relation by multiplying both sides of the above equation by λ b. Then we take the trace and use Eq on the RHS. After rearranging the result, we obtain A a µ = 2 Tr(λ a A µ ). (2.48) Note that the coefficient A a µ is a scalar. This is the analogue of the vector potential in electromagnetism. We use the definition of gauge transformation in Eq A µ A µ = UA µ U 1 i g 0 ( µ U) U 1 (2.49) and proceed to analyze the behavior of the field-strength tensor under gauge transformations. Let us try to use the Abelian-case definition of F µν in Eq Noticing the property µ U 1 = U 1 ( µ U) U 1, (2.50) which can be easily proven from µ (UU 1 ) = 0, we obtain [ F µν = µ UA ν U 1 i ] [ ( ν U)U 1 ν UA µ U 1 i ] ( µ U)U 1 g 0 g 0 = U ( µ A ν ν A µ ) U 1 + [ ( µ U) A ν ( ν U) A µ ] U 1 + U [ A ν ( µ U 1) A µ ( ν U 1)] i g 0 [ ( ν U) ( µ U 1 ) ( µ U) ( ν U 1 ) ] = UF µν U 1 + [ ( µ U)(U 1 U)A ν ( ν U)(U 1 U)A µ ] U 1 U [ A ν U 1 ( µ U) U 1 A µ U 1 ( ν U) U 1] + i g 0 [ ( ν U) U 1 ( µ U) U 1 ( µ U) U 1 ( ν U) U 1]

41 2.4 Non-Abelian Theory 39 = U 1 F µν U [ UA ν U 1, ( µ U) U 1] [ ( ν U) U 1, UA µ U 1] + i g 0 [ ( ν U) U 1, ( µ U) U 1], (2.51) where we introduced the commutator in the last step. Such a transformation will induce extra terms with dependence on A µ, changing the final equations of motion. We need thus to redefine our procedure, but we wish to keep our definition of a gauge transformation in Eq Note that if we calculate [A µ, A ν] we obtain [ ] A µ, A ν = [UA µ U 1 i ( µ U) U 1, UA ν U 1 i ] ( ν U) U 1 g 0 g 0 = [ UA µ U 1, UA ν U 1] + 1 g 2 0 [ ( ν U) U 1, ( µ U) U 1] + i g 0 [ UAν U 1, ( µ U) U 1] + i g 0 [ ( ν U) U 1, UA µ U 1]. (2.52) In the above expression, we notice the appearance of several terms that are in common with Eq Taking this into account, let us rewrite the transformation of F µν as defined in Eq F µν F µν = UF µν U 1 ig 0 [UA µ U 1, UA ν U 1 ] + ig 0 [A µ, A ν], (2.53) or, equivalently, ( ) F µν ig 0 [A µ, A ν] = U F µν ig 0 [A µ, A ν ] U 1. (2.54) as This hints at a natural definition of the field-strength tensor in the non-abelian case F µν = µ A ν ν A µ ig 0 [A µ, A ν ]. (2.55) Using this definition and recalling the transformation in Eq. 2.41, the result is that the field-strength tensor F µν behaves under gauge transformations as F µν F µν = UF µν U 1. (2.56)

42 40 Overview of Gauge Field Theories We see that, contrary to what happened in the Abelian case, the field-strength tensor is not gauge-invariant. Note that F µν has a matrix nature as well and can be decomposed in terms of the generators just as was done with the vector potential F µν = F c µνλ c, (2.57) with F c µν given by F c µν = 2 Tr(λ c F µν ) = µ A c ν ν A c µ + g 0 f abc A a µ A b ν, (2.58) where we used Eqs. 2.55, 2.46 and It is natural to extend this notation to the current j µ, i.e. j µ = j µ, a λ a, j µ, a = 2 Tr(λ a j µ ). (2.59) As was done for the Abelian theory, we will treat j µ, a for now as an external current, which can be useful for calculating correlation functions of the field A a µ. It will gain a physical interpretation as the current of color-charged fermions in Section 2.5. Now, having the Lagrangian density of electromagnetism in Eq as motivation, we write the Lagrangian density for the non-abelian theory as L = 1 4 F a µν F µν, a j µ, a A a µ. (2.60) The above expression can be written in matrix notation by inserting a Kronecker delta in each term. L = 1 4 F a µν δ ab F µν, b j µ, a δ ab A b µ = 1 2 F a µν F µν, b Tr(λ a λ b ) 2 j µ, a A b µ Tr(λ a λ b ) = 1 2 Tr(F µνf µν ) 2 Tr(j µ A µ ), (2.61) where we used the normalization in Eq

43 2.4 Non-Abelian Theory 41 We remark that j µ = j µ, a λ a is an external current and therefore we may choose its transformation to keep the terms involving it gauge-invariant. In any case, as in the Abelian theory, we will be able to interpret j µ physically in Section 2.5, obtaining that Tr(j µ A µ ) is not invariant under gauge transformations d. As happened before, it turns out that this extra term it is needed to keep the full Lagrangian invariant under gauge transformation. In this way, the Lagrangian is left invariant under gauge transformations L = 1 2 Tr ( F µνf µν ) 2 Tr ( ) j µ A µ = 1 2 Tr ( UF µν U 1 UF µν U 1) 2 Tr (j µ A µ ) = 1 2 Tr (F µνf µν ) 2 Tr (j µ A µ ) = L. (2.62) Finally, we may derive the equations of motion. The Euler-Lagrange equations are δl δa a ν δl µ δ( µ A a ν) = 0. (2.63) Let us substitute Eq into Eq and use δ ( ) j ρ, d A d ρ = j ν, c δa c ν (2.64) δfρτ d δa c ν δ ( ) j ρ, d A d ρ δ( µ A c ν) = g 0 f abd ( δ ν ρδ ac A b τ + δ ν τ δ bc A a ρ) = g0 f cbd A b τδ ν ρ + g 0 f acd A a ρδ ν τ (2.65) = 0 (2.66) δf d ρτ δ( µ A c ν) = δµ ρ δ ν τ δ cd δ ν ρδ µ τ δ cd (2.67) along with its similar versions for F µν, c, which can be obtained through contractions with the metric tensor (see Appendix A), to obtain j ν, c = 1 4 ( ) g0 f cbd A b τδρ ν + g 0 f acd A a ρδτ ν F ρτ, d 1 ( g0 f cbd A τ, b δ ρν + g 0 f acd A ρ, a δ τν) Fρτ d 4 d Note that if we choose j µ = Uj µ U 1, then the term containing the current will transform as Tr(j µ A µ ) Tr(j µ A µ ) i g 0 Tr(j µ U 1 µ U). Notice that, again, this extra term does not affect the equations of motion (see footnote in p. 37)and, for the moment, there is no impact in choosing Tr(j µ A µ ) to be invariant under gauge transformations.